MERMAIDS: Structured Memory and Evidence Reuse for Reducing Multi-Hop Retrieval Hallucinations in Agentic RAG

MERMAIDS: Structured Memory and Evidence Reuse for Reducing Multi-Hop Retrieval Hallucinations in Agentic RAG | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article MERMAIDS: Structured Memory and Evidence Reuse for Reducing Multi-Hop Retrieval Hallucinations in Agentic RAG Wenyou Huang, Hong Zhuang, Wen Huang, Junnan Kou, Ruoxuan Wei, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9455786/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Agentic Retrieval–Augmented Generation (RAG) systems have advanced the ability of large language models to handle complex, multi–step questions by dynamically planning and executing retrieval operations. However, these systems remain vulnerable to intermediate reasoning breakdowns, redundant retrieval, and evidence inconsistency, which collectively cause answer drift in multi–hop question answering. We propose MERMAIDS, a framework that augments Agentic RAG with a Structured Evidence Memory (SEM) module and a Cross-task Evidence Reuse Cache (ERC). Extracted evidence is organized into a lightweight knowledge graph following a claim–evidence–source–time schema, and a dedicated Conflict Detection module identifies contradictions and triggers selective re-retrieval or answer confidence degradation. Experiments on HotpotQA, MuSiQue, MultiHop-RAG, and FEVER show that MERMAIDS achieves 5.9–8.6 point improvements in exact match accuracy over strong baselines, while reducing redundant retrieval steps by 28–35% and attaining conflict detection precision above 81%. Retrieval-Augmented Generation Agentic AI Multi-Hop Reasoning Knowledge Graph Evidence Conflict Detection Hallucination Mitigation Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9455786","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":625378043,"identity":"48f4befd-e1c5-4e06-ba1a-c07a09115e5f","order_by":0,"name":"Wenyou Huang","email":"","orcid":"","institution":"Stevens Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Wenyou","middleName":"","lastName":"Huang","suffix":""},{"id":625378044,"identity":"5b33da40-43a5-435c-92be-22f3d9ae9099","order_by":1,"name":"Hong Zhuang","email":"","orcid":"","institution":"Hong Zhuang University of Illinois at Urbana Champagne","correspondingAuthor":false,"prefix":"","firstName":"Hong","middleName":"","lastName":"Zhuang","suffix":""},{"id":625378045,"identity":"cdaf3f00-fdde-4b81-bcc7-22d6508a216b","order_by":2,"name":"Wen Huang","email":"","orcid":"","institution":"Independent Researcher","correspondingAuthor":false,"prefix":"","firstName":"Wen","middleName":"","lastName":"Huang","suffix":""},{"id":625378046,"identity":"7b1ffec6-c8ad-427e-b29d-ab154c389a08","order_by":3,"name":"Junnan Kou","email":"","orcid":"","institution":"University of Washington","correspondingAuthor":false,"prefix":"","firstName":"Junnan","middleName":"","lastName":"Kou","suffix":""},{"id":625378047,"identity":"efb43396-8125-40f1-a43b-0b32784384df","order_by":4,"name":"Ruoxuan Wei","email":"","orcid":"","institution":"Trine University","correspondingAuthor":false,"prefix":"","firstName":"Ruoxuan","middleName":"","lastName":"Wei","suffix":""},{"id":625378048,"identity":"e7a70073-1912-4cf3-85e8-06839dfeb60c","order_by":5,"name":"Ning Lyu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA70lEQVRIiWNgGAWjYDACZiB+wMAgx8befBBIM/DwEaUlgYHBmJ/nWLIBSAsbUTYBtSTOnJGjJgHiENRizs57+EVi22HGDQdy2Cq/5tjJsDEwP3x0A48Wy2a+NAugFmaDA2eP3Zbdlgx0GJuxcQ4eLQaHecwMgFrYDA72pd2W3MYM1MLDJk2MFh4Qo1hyWz1RWowfALVISLbxmDF+3HaYOFsYEs6lG/DzsCVLM247zsPGTMgv588Yf/hQZl3fJv/44Mef26rt+dmbHz7GpwUI2CQYoXHBzAMm8SsHK/nA8AfCYvxBWPUoGAWjYBSMQAAA7qVHbz4ryYoAAAAASUVORK5CYII=","orcid":"","institution":"Carnegie Mellon University","correspondingAuthor":true,"prefix":"","firstName":"Ning","middleName":"","lastName":"Lyu","suffix":""}],"badges":[],"createdAt":"2026-04-18 09:48:55","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9455786/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9455786/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107868301,"identity":"9f1a0320-8566-4e4f-83f5-d53c0eb99b4a","added_by":"auto","created_at":"2026-04-27 07:09:53","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":488658,"visible":true,"origin":"","legend":"","description":"","filename":"StructuredMemoryandEvidenceReuseforReducingMultiHopRetrievalHallucinationsinAgenticRAGcopy.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9455786/v1_covered_f5f1fc78-8cee-4b5d-a525-f8adf49c67c3.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eMERMAIDS: Structured Memory and Evidence Reuse for Reducing Multi-Hop Retrieval Hallucinations in Agentic RAG\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Retrieval-Augmented Generation, Agentic AI, Multi-Hop Reasoning, Knowledge Graph, Evidence Conflict Detection, Hallucination Mitigation","lastPublishedDoi":"10.21203/rs.3.rs-9455786/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9455786/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAgentic Retrieval–Augmented Generation (RAG) systems have advanced the ability of large language models to handle complex, multi–step questions by dynamically planning and executing retrieval operations. However, these systems remain vulnerable to intermediate reasoning breakdowns, redundant retrieval, and evidence inconsistency, which collectively cause answer drift in multi–hop question answering. We propose MERMAIDS, a framework that augments Agentic RAG with a \u003cem\u003eStructured Evidence Memory\u003c/em\u003e (SEM) module and a \u003cem\u003eCross-task Evidence Reuse Cache\u003c/em\u003e (ERC). Extracted evidence is organized into a lightweight knowledge graph following a claim–evidence–source–time schema, and a dedicated \u003cem\u003eConflict Detection\u003c/em\u003emodule identifies contradictions and triggers selective re-retrieval or answer confidence degradation. Experiments on HotpotQA, MuSiQue, MultiHop-RAG, and FEVER show that MERMAIDS achieves 5.9–8.6 point improvements in exact match accuracy over strong baselines, while reducing redundant retrieval steps by 28–35% and attaining conflict detection precision above 81%.\u003c/p\u003e","manuscriptTitle":"MERMAIDS: Structured Memory and Evidence Reuse for Reducing Multi-Hop Retrieval Hallucinations in Agentic RAG","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-21 11:33:03","doi":"10.21203/rs.3.rs-9455786/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e1257db3-5000-410e-94e5-b4ce2cdfc873","owner":[],"postedDate":"April 21st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-21T11:33:03+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-21 11:33:03","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9455786","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9455786","identity":"rs-9455786","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}